Rxivist logo

A new colorectal cancer risk prediction model incorporating family history, personal and environmental factors

By Yingye Zheng, Xinwei Hua, Aung Ko Win, Robert J. MacInnis, Steven Gallinger, Loic Le Marchand, Noralane M. Lindor, John A. Baron, John L Hopper, James G. Dowty, Antonis C. Antoniou, Jiayin Zheng, Mark A. Jenkins, Polly A. Newcomb

Posted 07 Jun 2019
bioRxiv DOI: 10.1101/662106

Purpose: Reducing colorectal cancer (CRC) incidence and mortality through early detection would improve efficacy if targeted. A CRC risk-prediction model incorporating personal, family, genetic and environmental risk factors could enhance prediction. Methods: We developed risk-prediction models using population-based CRC cases (N=4,445) and controls (N=3,967) recruited by the Colon Cancer Family Registry Cohort (CCFRC). A familial risk profile (FRP) was calculated to summarize individuals risk based on their CRC family history, family structure, germline mutation probability in major susceptibility genes, and a polygenic component. Using logistic regression, we developed risk models including individuals FRP or a binary CRC family-history (FH), and risk factors collected at recruitment. Model validation used follow-up data for population- (N=12,052) and clinic-based (N=5,584) relatives with no cancer history at recruitment, assessing calibration (E/O) and discrimination (AUC). Results: The E/O (95% confidence interval [CI]) for FRP models for population-based relatives were 1.04 (0.74-1.45) and 0.86 (0.64-1.20) for men and women, and for clinic-based relatives 1.15 (0.87-1.58) and 1.04 (0.76-1.45). The age-adjusted AUC (95% CI) for FRP models in population-based relatives were 0.69 (0.60-0.78) and 0.70 (0.62-0.77), and for clinic-based relatives 0.77 (0.69-0.84) and 0.68 (0.60-0.76). The incremental values of AUC (95% CI) for FRP over FH models for population-based relatives were 0.08 (0.01-0.15) and 0.10 (0.04-0.16), and for clinic-based relatives 0.11 (0.05-0.17) and 0.11 (0.06-0.17). Conclusion: The FRP-based model and FH-based model calibrate well in both settings. The FRP-based model provided better risk-prediction and discrimination than the FH-based model. A detailed family history may be useful for targeted risk-based screening and clinical management.

Download data

  • Downloaded 269 times
  • Download rankings, all-time:
    • Site-wide: 94,999
    • In epidemiology: 3,912
  • Year to date:
    • Site-wide: 117,022
  • Since beginning of last month:
    • Site-wide: 91,003

Altmetric data

Downloads over time

Distribution of downloads per paper, site-wide